Our test bench: about thirty engines measured on the same archives. See the engines compared →
| Tool | Who makes it | Where | Status · model | What it does | What it doesn’t do |
|---|---|---|---|---|---|
| wondarchive | A researcher, for their own fonds first | 🇫🇷 France | company · plans and extra pages | deposit, straightening, transcription, translation, reading, search, shelfmarks, correction, tags, notes, timeline and map; one-click citation and TEI coming soon | training a model on a specific hand |
| Transkribus | READ-COOP SCEEuropean cooperative founded in 2019, 250 member institutions | 🇦🇹 Innsbruck | cooperative · subscription and credits (Scholar €99/year, 900 pages) | handwriting and print recognition, trainable models, tables, publishing sites | translation, managing trip photos, straightening |
| eScriptorium and Kraken | EPHE – Université PSLScripta project, digital humanities team; Kraken by Benjamin Kiessling | 🇫🇷 Paris | public research · free, open-source software | very fine segmentation and recognition, model training | hosting — needs a server and an engineer; translation; a general-public interface |
| Arkindex | Tekliaspecialist archives-and-museums SME, since 2019 | 🇫🇷 Paris | company · open platform, services for institutions | mass processing, document pipelines, structured extraction, API | selling to individuals: it’s an institutional tool |
| Tropy | RRCHNM, George Mason University, with C²DH Luxembourg and Digital Scholar | 🇺🇸 Virginia · 🇱🇺 | university · free, open-source software, on your machine | organising, describing and annotating archive photos | no transcription, no translation, nothing online |
| ABBYY FineReader | ABBYYpublisher, United States | 🇺🇸 | company · software licence | good-quality desktop OCR on clean print, PDF | crooked trip photos, manuscript, a corpus, translation |
| Google Document AI · Amazon Textract · Azure AI Vision | Google, Amazon, Microsoft | 🇺🇸 | cloud platforms · pay-per-call, for developers | robust generic OCR, to integrate yourself | everything else: no interface, no corpus, no reading |
| OCR4all | University of Würzburg | 🇩🇪 | public research · free, open-source | old printed matter, Gothic script, full pipeline to install | hosting, translation, trip photos |
| DeepL | DeepL SE | 🇩🇪 Cologne | company · subscription | translation only, very high quality | doesn’t read an image; it’s the next step, not the tool |
Facts checked in September 2026 on publishers’ own sites: READ-COOP SCE, eScriptorium at EPHE, Arkindex by Teklia, Tropy at RRCHNM. Transkribus prices are those shown on their own site.
Which tool for your fonds?
Reading-room photos, manuscript, Fraktur, clean scans, no budget or a lab: the guide to choosing starts from your situation and ends with a tool.
A comparison per tool
- wondarchive or Transkribusthe main duel: trainable manuscript against reading, translation and collection management
- wondarchive or eScriptoriumopen source to install, or a service ready to go
- wondarchive or Tropythey organise the photos, we read them
- wondarchive or ABBYY FineReaderdesktop OCR against reading-room photos
- wondarchive or cloud OCRpowerful, to integrate yourself
- wondarchive or Arkindextwo French tools, two audiences
See it on your own pages
A hundred pages transcribed and translated, for a year, no card required.